Discover more from Last Week in AI
Last Week in AI #194: Amazon's new warehouse robots, Meta takes down language model demo after 3 days, anime convention bans AI art, and more!
Amazon's warehouse robots can now stow items into storage pods, Meta takes down Galactica language model demo after 3 days in response to controversies, Anime LA bans AI art from displays and booths
Amazon unveiled its latest robotics project that aims to automate the labor-intensive stowing operation in its fulfillment centers. Stowing is the process of putting individual items into small compartments of a large storage pod. This task is much more challenging than the typical pick-and-place operations done by many logistics robot startups. It requires the robot to manipulate objects in dense clutter, reason about forces and contacts with the environment, and handle a wide variety of objects. Amazon execs were convinced by this project’s utility near the end of summer 2021, and since then the team has been pushing hard at making the system deployable across Amazon’s fulfillment centers. The eventual goal is the ability to stow 85% of Amazon’s products.
Our take: This type of contact-rich robot manipulation with novel objects in clutter is really challenging, and it’s super exciting that Amazon seemingly has a system that achieves this capability in its fulfillment centers. Undoubtedly, a lot of recent advances in AI and robotics were needed to bring this to fruition, from robust object segmentation algorithms to robust contact-rich robot motion planners. This is also another example of how the most impactful application of AI and robotics may not be something that the end-user directly sees, like robot butlers, but more through indirect domains, like automation in logistics that brings down the cost of e-commerce.
On November 15, Meta released the latest in a line of large language models. Galactica had been trained on 48 million examples of scientific documents including articles, textbooks, and encyclopedias. The promise was huge: Galactica was introduced as a model that "can summarize academic literature, solve math problems, generate Wiki articles, write scientific code, annotate molecules and proteins, and more." But the model couldn't quite live up to its hype, as some researchers found. After just three days, the Galactica demo was taken down
Our Take: There's been a lot of discourse here--it is indeed troubling that Galactica spat out wrong or biased information with an authoritative tone. This pushback is deserved: Galactica unsurprisingly inherits the faults of its predecessors. The underlying question seems to be the same as it always was: "Should these models be released in the first place?" I do think there is something to the idea that LLMs can improve the way we access and engage with information, but LLMs will need to be substantially better than they are today to do this effectively. Does that mean we should shut down open demos until then? Or is there a different model that can encourage progress without the large risks that accompany a company like Meta's choice to release Galactica to a large number of people? I think iterations and mistakes are needed to move the field forward and allow us to determine how language models and AI systems can help us do things better. But making the same mistake over and over again isn't progress. We've seen Galactica's problems before. Meta should be able to do better.
AI-generated artwork has increasingly been the center of controversy in recent years. Animé Los Angeles, an anime convention, has now issued a statement concerning AI-generated art and clarified its policy. The convention does not allow AI-generated art to be displayed or sold in its exhibit halls and Artist Alleys. The statement also makes clear that using AI-generated art in promotional materials is prohibited and that selling AI-generated art is considered counterfeit/bootleg merchandise and will be required to be removed.
Our take: Legal, ethical, and normative use of AI-generated art is still very much up in the air, and acts like these may inform the future norms regarding both the commercial and hobbyist applications of AI art. It is not surprising that large organizations are taking the least risky approach - banning AI art altogether - instead of going for something more nuanced that may incur controversies. It is likely that we will see similar bans in the near future until clear legal precedants have been established, either by legislation or lawsuits.
The State of Multilingual AI - "Models that allow interaction via natural language have become ubiquitious. Research models such as BERT and T5 have become much more accessible while the latest generation of language and multi-modal models are demonstrating increasingly powerful capabilities."
We’re getting a better idea of AI’s true carbon footprint - "Large language models (LLMs) have a dirty secret: they require vast amounts of energy to train and run. What’s more, it’s still a bit of a mystery exactly how big these models’ carbon footprints really are."
Solving brain dynamics gives rise to flexible machine-learning models - "MIT CSAIL researchers solve a differential equation behind the interaction of two neurons through synapses to unlock a new type of speedy and efficient AI algorithm."
MIT solved a century-old differential equation to break 'liquid' AI's computational bottleneck - "Last year, MIT developed an AI/ML algorithm capable of learning and adapting to new information while on the job, not just during its initial training phase."
A review of machine learning-based failure management in optical networks - "In the present big data era, data is growing exponentially, and optical networks serve as the backbone for high-capacity and long-distance data transmission. Optical networks are always large scale, comprise massive components, and cover a wide area."
Developing robust benchmarks for driving forward AI innovation in healthcare - "Machine learning technologies have seen increased application to the healthcare domain. The main drivers are openly available healthcare datasets, and a general interest from the community to use its powers for knowledge discovery and technological advancements in this more conservative field."
Language Models are Changing AI: The Need for Holistic Evaluation - "We benchmark 30 prominent language models across a wide range of scenarios and for a broad range of metrics to elucidate their capabilities and risks."
AI-generated x-ray images fooled medical experts and improved osteoarthritis classification - "Sharing medical data between laboratories and medical experts is important for medical research. However, data sharing is often sufficiently complex and sometimes even impossible due to the strict data regulatory legislation in Europe."
AI’s new frontier: Connecting grieving loved ones with the deceased - "How scientists are using virtual reality to create avatars, chatbots and even eternal digital entities."
This Comic Series Is Gorgeous. You'd Never Know AI Drew the Whole Thing - "The Bestiary Chronicles is both a modern fable on the rise of artificial intelligence and a testament to how shockingly fast AI is evolving."
The Deepfake Music Industry is Coming - "Generative AI will transform Creative Industries, or will they?"
Sniper robot treats 500k plants per hour with 95% less chemicals | Challengers - "This “intelligent sharpshooter” farming robot distinguishes crops from weeds — and it could help feed 10 billion people."
Protein programmers get a helping hand from Cradle’s generative AI - "Proteins are the molecules that get work done in nature, and there’s a whole industry emerging around successfully modifying and manufacturing them for various uses."
Autonomous robots to help modernize grape, wine industry - "Next spring, a small army of Cornell-developed PhytoPatholoBots (PPB) will be deployed to four grape breeding programs across the U.S. on a mission to guide the global grape and wine industry into the 21st century."
Using machine learning to assess the livelihood impact of electricity access - "In many regions of the world, sparse data on key economic outcomes inhibit the development, targeting and evaluation of public policy1,2. We demonstrate how advancements in satellite imagery and machine learning (ML) can help ameliorate these data and inference challenges."
Intel's new deepfake detector can spot a real or fake video based on blood flow in video pixels - "Intel's FakeCatcher detects a deepfake in real time with a 96% accuracy rate"
MoMA’s newest artist is an AI trained on 180,000 works, from Warhol to Pac-Man - "Unsupervised reaches new highs in what’s possible with AI, creating a breathing being that synthesizes 180,000 artworks from MoMA."
AI Time Machine Allows You to Picture Yourself in any Historical Period - "AI Time Machine is a new tool that allows users to create images of a person in different time periods throughout history using AI-image generator technology."
Notion is using AI to automatically write your blog posts, job descriptions, and poetry - "Someday soon, Notion will be able to write your notes for you. The company is starting to test a new feature today called Notion AI, which Notion CEO Ivan Zhao tells me could ultimately change the way people use the app — and do their jobs."
HLTH22: Nuance, Nvidia put AI tools in the hands of clinicians, opening pipeline 'from bench to bedside' - "Nuance Communications, a Microsoft company, and Nvidia today announced a partnership they say will open the pipeline between AI technology and real-world applications."
Sony Patents Machine Learning System to Show Spectators Button Inputs - "Sony is exploring a controller input recognition technology that could significantly enhance the experience of watching gameplay by providing viewers with accurate controller input captions inferred from game footage."
ANALYSIS: Patents Forecast Widespread Reach of AI Tech in 2023 - "Artificial intelligence is driving important developments in technology, from controlling autonomous vehicles, to developing medical diagnoses, to combating climate change. The global AI market was valued at nearly $59.7 billion in 2021, and is estimated to reach $422.4 billion by 2028."
Cerebras Reveals Andromeda, a 13.5 Million Core AI Supercomputer - "Cerebras, the company that builds the world's largest chip, the Wafer Scale Engine 2 (WSE-2), unveiled its Andromeda supercomputer today. Andromeda combines 16 of the wafer-sized WSE-2 chips into one cluster with 13."
Boston Dynamics suing Ghost Robotics over quadruped design - "Boston Dynamics is suing competitor Ghost Robotics for patent infringement. Boston Dynamics claims Ghost Robotics infringed on seven patents related to its Spot quadruped robot."
Report finds employees embrace AI when they see its value - "For organizations to realize value from artificial intelligence, individual employees must embrace AI and clearly see its benefits."
KKR Leads US$68 million Series B Funding Round in AI Leader Advanced Navigation - "Advanced Navigation (the "Company"), a developer of AI robotics and navigation technology, and KKR, a leading global investment firm, today announced the signing of definitiv"
Owl Labs raises $25 million in Series C funding round led by HP Tech Ventures - "Today, we are proud to announce a $25 million Series C investment and a strategic partnership with HP, bringing our company’s total funding to over $47 million. The funding round was led by HP Tech Ventures, the venture capital arm of HP Inc."
How DeviantArt is navigating the AI art minefield - "This is not something that DeviantArt can fix on our own.’"
AI-powered media editing app Descript lands fresh cash from OpenAI - "Descript, the audio and video editing platform founded in 2017 by former Groupon CEO Andrew Mason, has raised $50 million in a Series C round led by the OpenAI Startup Fund, a tranche through which OpenAI and its partners, including Microsoft, are investing in early-stage companies."
NVIDIA Teams With Microsoft to Build Massive Cloud AI Computer - "NVIDIA today announced a multi-year collaboration with Microsoft to build one of the most powerful AI supercomputers in the world"
Accrete wins contract from Pentagon for AI threat detection software - "Accrete said it won a five-year, multi-million dollar Production Operational Technology software licensing contract from the U.S. Department of Defense for Argus, the company’s threat detecting AI software."
Spot AI raises $40M to build smarter CCTV security camera tech - "CCTV and other kinds of security cameras have a strong big brother vibe, but for many of us that may be because we don’t really understand or know how the footage they pick up ever gets used."
Ubisoft and Riot are going to use AI to stop you from being horrible online - "Toxic gamers who enjoy harassing others via in-game comms are set for an equally rude awakening. As creators and publishers of some of the best co-op games out there, Ubisoft and Riot Games are no strangers to having to deal with toxic players."
UAE Royals Back AI Startup Founded by Former Goldman Banker - "The Applied AI Co. draws funding from G42, Dubai royals"
Qualcomm Teases Powerful New Snapdragon Laptop CPU, Claims AI Processing Will Redefine PCs - "Windows computers have been available with ARM-based chips for years, but the OS still doesn't work as optimally as it does on x86. That hasn't stopped Qualcomm from rolling out ever more powerful laptop chips like the current-gen Snapdragon 8cx Gen 3."
Making data available on demand for AI and ML workloads, Weka lands $135M - "As global enterprises continue to become bullish on data-driven applications, the companies providing the backend for these apps are seeing a massive surge in demand and investor support. "
AEYE Health gets FDA clearance to use AI to screen diabetics in hopes of preventing blindness - "That clearance could be a game-changer for the millions of people who are at high risk of diabetic retinopathy."
Tesla Tells US of New Fatal Crashes Involving Automated Driving - "Tesla Inc. disclosed to US regulators two new fatal crashes involving automated driver-assist systems, bringing its total to 16 since the government required carmakers to begin submitting data on such accidents in June 2021."
‘Full-on robot writing’: the artificial intelligence challenge facing universities - "AI is becoming more sophisticated, and some say capable of writing academic essays. But at what point does the intrusion of AI constitute cheating?"
What Riding in a Self-Driving Tesla Tells Us About the Future of Autonomy - "Cade and Ian spent six hours riding in a self-driving car in Jacksonville, Fla., to report this story."
The scary truth about AI copyright is nobody knows what will happen next - "Generative AI has had a very good year. Corporations like Microsoft, Adobe, and GitHub are integrating the tech into their products; startups are raising hundreds of millions to compete with them; and the software even has cultural clout, with text-to-image AI models spawning countless memes."
To Regulate General Purpose AI, Make the Model Move - "The EU has brought regulating general-purpose AI (GPAI) models, like GPT-3 and Stable Diffusion, to the forefront of the AI policy debate. "
Can AI-generated improvised comedy be funny? - BBC News - "Artificial intelligence (AI) is being used in live comedy shows to make people laugh – so how does it work?"
I record myself on audio 24x7 and use an AI to process the information. Is this the future? - "What will happen when our phones record everything we say 24x7 using AI to process that information?, before someone else does, I tried to do it myself, and this is the result."
AI Can Now Make Fake Selfies For Your Tinder Profile - "The AI image-generating craze has entered its next phase of absurdity: creating fake profile pics that make you look good on dating apps and social media."
Copyright © 2022 Skynet Today, All rights reserved.